基于多源图像识别技术的信号设备室智慧巡检系统

Intelligent Inspection System of Signal Equipment Room Based on Multi-source Image Recognition Technology

针对城市轨道交通信号设备发生故障或隐患时,难以通过日常巡检及时发现的问题,提出基于多源图像识别技术的信号设备室智慧巡检系统。采用YOLOv5目标检测算法实现设备指示灯颜色及状态的自动识别,通过对算法模型进行训练和分析,将识别结果匹配设备信息,同时诊断设备运行状态,自动生成设备巡检报告和异常报告。在广州地铁3号线的试点应用中,验证分析了设备指示灯识别准确率与像素值的关系,以及在满足监控覆盖时的摄像头布设方案。实验结果表明,基于多源图像识别技术的设备室智慧巡检系统巡检效率高,漏检率低,能够大幅减少人工巡检成本,具有广阔的市场应用前景和较高的应用价值,对强化城市轨道交通智能运维具有重要意义。

Considering that it is difficult to identify problems in time through the daily inspection in case of failure or hidden danger of the urban rail transit signal equipment,an intelligent inspection system of the signal equipment room based on multi-source image recognition technology is presented.The system applies the YOLOv5 object detection algorithm to achieve automatic recognition of device indicator light color and on/off status.Through training and analysis of the algorithm model,the identification results will be matched with the device information and the diagnosis of the operating status of the device will be made,and the device inspection report and anomaly report will be automatically generated.In the pilot application of Guangzhou Metro Line 3,the relationship between device indicator recognition accuracy and pixel value is analyzed and verified,including the camera layout scheme which can satisfy the requirement for monitoring coverage.The results show that the intelligent inspection system of signal equipment room based on multi-source image recognition technology features high efficiency and low omission rate,which can greatly reduce the manual inspection cost.With high application value,the system has a broad market prospect and is of great significance to strengthen the intelligent operation and maintenance of urban rail transit.